llama381binstruct_summarize_short

This model is a fine-tuned version of NousResearch/Meta-Llama-3.1-8B-Instruct on the generator dataset. It achieves the following results on the evaluation set:

• Loss: 2.7961

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 30
• training_steps: 500

Training results

Training Loss	Epoch	Step	Validation Loss
1.5831	1.25	25	1.6479
0.8086	2.5	50	1.6711
0.3333	3.75	75	1.8202
0.1834	5.0	100	1.9703
0.083	6.25	125	2.2035
0.0381	7.5	150	2.3948
0.0212	8.75	175	2.4440
0.0197	10.0	200	2.4377
0.0188	11.25	225	2.3988
0.0094	12.5	250	2.5389
0.0062	13.75	275	2.5661
0.0048	15.0	300	2.6391
0.0036	16.25	325	2.6828
0.0021	17.5	350	2.7307
0.0017	18.75	375	2.7560
0.0023	20.0	400	2.7722
0.0024	21.25	425	2.7839
0.002	22.5	450	2.7908
0.0021	23.75	475	2.7946
0.0018	25.0	500	2.7961

Framework versions

• PEFT 0.12.0
• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1